clickmax-temperature-score
Use when the user wants to understand, tune, or audit contact Temperature (behavioural 0-100 engagement) and contact Score (criteria-based fit points) in Clickmax, including why one contact has a given number and the Temperature × Score health of the base.
How do I install this agent skill?
npx skills add https://docs.clickmax.io --skill clickmax-temperature-scoreIs this agent skill safe to install?
No partner audit is available yet. Read the source before installing.
What does this agent skill do?
As tools abaixo aparecem com os nomes que o MCP da Clickmax registra. Se o seu cliente de IA prefixar nomes de tool (
mcp__<servidor>__,mcp_<servidor>_, ou outro), use o nome já prefixado que aparecer na sua lista de tools.
When this applies
Use for the two per-contact indicators of the CRM: Temperature (what the contact DID, 0-100, decays) and Score (who the contact IS, points from rules, does not decay). Covers reading/changing their configuration, explaining one contact's number, and measuring the base.
Not this skill:
- filtering/listing contacts by number →
clickmax-leads,clickmax-list-segments(fieldstemperatureScore,leadScore) - the hot/warm/cold/frozen label typed by hand on an opportunity card →
clickmax-pipelines - automations that react to or nudge Temperature →
clickmax-flows
Key assumptions
| Topic | Fact |
|---|---|
| Temperature input | behaviour only: messages, e-mail opens/clicks, pages/checkout, payments, members area, meetings, pipeline events; registration fields never count |
| Temperature model | 4 sub-scores 0-100: Recency, Frequency, Depth, Declaration (intent to buy); blended by the active model; R and F decay (halflifeR/F days), P and D do not decay inside windowDays |
| Bands | cold < thresholdMorno ≤ warm < thresholdQuente ≤ hot; default preset lancamento = 30 / 60 |
| Model vs numbers | activeModelId only picks the R/F/P/D blend. Thresholds, channel multipliers, window and half-lives are the numbers saved in settings: switching model does NOT copy that model's numbers |
| Per-event weights | eventWeights holds only overrides (−100..100); absent event = product default; channel multiplier 0 silences the whole channel; the field replaces the stored map when sent |
| Score formula | sum of points of ENABLED rules the contact satisfies now, floor 0, cap maxScore; tier low/medium/high by thresholdMedium/thresholdHigh |
| Score default | OFF. Rules can exist but nobody gets a number until enabled: true; turning it off clears every stored Score |
| Rule criteria | same filter tree as a segment; describes the contact (data, tags, origin, purchases), never behaviour; empty criteria scores nobody |
| Blast radius | any settings/rule save re-scores the whole base asynchronously (seconds to minutes); the write returns the saved config, not the new numbers |
| Automations | a Temperature settings save can make contacts change band and fire temperature-triggered automations in bulk; a Score settings/rule re-score fires nothing |
| Permissions | reads = any role; writes = workspace admin/editor or coordinator attendant |
| Previews | model switch → temperature_model_impact (bands now vs after, ≤400 active contacts re-scored then projected, inactive exact, no hysteresis, uses the TARGET model's own numbers); Score rule → lead_score_rules_preview with points (+ ruleId when editing) = net tier delta from the stored, already clamped Score — approximations, the re-score after saving is the truth |
| Not exposed | the six preset models are read-only; no tool enumerates the event catalog (names come from temperature_lead_explain); the per-slider live preview of the settings screen is browser-side only |
Thought process
- Classify the ask: understand one contact | audit the base | change configuration.
- Understand/audit first, change second. A change never precedes reading current settings, because writes are full replaces.
- Temperature question ("is engaged?") vs Score question ("is a good fit?") → never answer one with the other's tool.
- Config changes are workspace-wide and slow to show: state that, then verify with a read after a short wait.
Execute guide
- Why this contact? Resolve the contact id (
clickmax-leads), then calltemperature_lead_explainfor Temperature andlead_score_lead_explainfor Score. Temperature: comparestoredvslive(they differ because R/F decay and the nightly pass only saves 3+ point drifts), name the top events byweightedPoints, and offermissingSignalsas what would raise it. Score: list rules bystatus;disabledshows what would have matched,clampedBysays if the floor/ceiling hid part of the sum. - State of the base.
lead_indicators_metricswithfilters = [](or a segment-style audience) gives bands, tiers, the matrix andpriority(hot AND high Score) in one call.lead_score_distributionis the Score-only cut.score.enabled = falseexplains an all-noneScore column. - Switch model ("what if").
temperature_model_impactwith the candidatemodelId→ comparecurrentvsnextper band (say "estimated" whenestimated: true). The preview assumes the target model's own thresholds/channels/window/half-lives; the switch only matches it iftemperature_settings_updatesends those same numbers with the newactiveModelId. - Tune Temperature.
temperature_settings_get→ change only what was asked →temperature_settings_updatewith EVERY scalar field resent. To adopt a preset wholesale, copy its numbers frombuiltInModelsinto the same call together withactiveModelId. To change one event weight, resend the whole storedeventWeightswith that entry edited ({}clears all overrides; omitting the field keeps them). - Saved calibration ("my model").
temperature_custom_models_createstores a snapshot only; it takes effect aftertemperature_settings_updatewith its id + its numbers. Update/delete affect just the snapshot; deleting the ACTIVE model is refused, and max 20 per workspace. - Set up Score.
lead_score_settings_get→ for each rule:lead_score_rules_previewwithpoints(size + tier shift + 3 examples) →lead_score_rules_create→ finallylead_score_settings_updatewithenabled: trueif it is still off (confirmthresholdHigh>thresholdMediumand ≤maxScore). - Pause/resume a rule.
lead_score_rules_set_enabledflips onlyenabled(re-scores the base). Before switching ON, preview with the rule'sfilters+points. Prefer it overlead_score_rules_delete, which is permanent. - Edit a rule. Preview the change with
lead_score_rules_previewpassing newfilters,pointsAND theruleId(so today's contribution is subtracted).lead_score_rules_updateis a full replace: omittedenabledbecomes true,descriptionnull,sortOrder0, criteria replaced. Read the rule first and resend everything that should stay. - Nudging one contact's Temperature from an automation is the flow action "Warm up or cool down contact" (signal: light warm, warm, cool, strong cool) →
clickmax-flows; it records a signal that fades like any activity, it does not write a number.
Report
- Lead with the indicator (Temperature or Score) and the job: explanation, base audit, or configuration change.
- Explanation: the number and band first, then at most 5 contributing events/rules ordered by impact (
+N more), then what would move it. Say "not calculated yet" for null, never 0. - Base audit: totals per band/tier, the
prioritycount, then the matrix cells with the largest counts. - Preview: bands/tiers before → after (or the net delta per tier) plus the examples; label it an estimate.
- Configuration change: the before → after of each changed field, then the blast-radius note (whole base re-scored, may take minutes; automations may fire for Temperature). Never quote new contact counts before a re-read confirms them.
- Follow-up mutations are opt-in only.
Warnings
- Temperature settings save can trigger automations in bulk; confirm with the user before saving and prefer small steps.
- Score rules and settings writes are silent no-ops for contacts while the Score is off.
- A stale
livevsstoredgap is normal decay, not a bug. - Do not put behaviour ("opened e-mail", "visited page") in a Score rule; it belongs to Temperature.
- Do not use the card hot/warm/cold label as if it were the contact Temperature number.
Anti-patterns
- Sending a partial settings body or a partial rule update (both are full replaces).
- Assuming a preset switch loaded that preset's thresholds and channel weights.
- Deleting a rule, or resending it whole through update, just to pause it.
- Switching model without sending the target model's numbers, then expecting the previewed bands.
- Creating rules and reporting success while the Score is still disabled.
- Reporting a re-scored count right after the write, before the background job finished.
Clickmax skill revision: 23f2b5440092
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/docs.clickmax.io/clickmax-temperature-score">View clickmax-temperature-score on skillZs</a>